EUVIMEDEUVIMED entdecken
Crossref

LUMIR: an LLM-driven unified agent framework for multi-task infrared spectroscopy reasoning

Zujie Xie, Zixuan Chen, Jiheng Liang, Xiangyang Yu, Ziru Yu

Analytica Chimica Acta · 2026-01 · 1382 · 344857

DOI: 10.1016/j.aca.2025.344857 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Analytica Chimica Acta, 2026-01. Verfasst von Zujie Xie und 4 weiteren Autor:innen. Beteiligte Einrichtung: School of Physics, State Key Laboratory of Optoelectronic Materials and Technologies, Sun Yat-Sen University, Guangzhou, 510275, China. Hauptthemen laut MEDLINE: Spectrophotometry, Infrared, Large Language Models. Die Arbeit wurde bisher 10-mal zitiert (Crossref) und nennt 43 Literaturangaben.

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Infrared spectroscopy enables rapid and non-destructive characterization of chemical and material properties, yet effective analysis typically requires workflows involving preprocessing, variable selection, and modeling. The construction and optimization of such workflows demand substantial expertise and often fail to generalize across datasets and tasks. Large language models (LLMs), with strong reasoning and adaptability, provide a promising approach to automate and enhance these processes. Nonetheless, existing chemometric and machine learning methods remain limited in robustness and transferability, leaving the field without a unified, data-efficient framework for automated infrared spectral analysis. This study introduces LUMIR (LLM-driven Unified agent framework for Multi-task Infrared spectroscopy Reasoning), an agent based framework designed to achieve accurate infrared spectral analysis under low data conditions. LUMIR integrates a structured literature knowledge base, automated preprocessing, feature extraction, and predictive modeling into a unified pipeline. By mining peer reviewed spectroscopy studies, it identifies validated preprocessing and feature derivation strategies, transforms spectra into low dimensional representations, and applies few-shot prompts for classification, regression, and anomaly detection. The framework was validated on diverse datasets, including the publicly available Milk near-infrared dataset, Chinese medicinal herbs, Citri Reticulatae Pericarpium(CRP) with different storage durations, an industrial wastewater COD dataset, and two additional public benchmarks, Tecator and Corn. Across these tasks, LUMIR achieved performance comparable to established machine learning and deep learning models, particularly in resource limited settings. This work demonstrates that by exploiting the broad generalization capacity of LLMs, together with structured literature guidance and few-shot learning, it is possible to achieve robust and automated spectral interpretation across diverse tasks. LUMIR establishes a new paradigm for applying LLMs to infrared spectroscopy, offering high accuracy with minimal labeled data and broad applicability across scientific and industrial domains.

Schlagwörter

MeSH-Schlagwörter (MEDLINE)

Spectrophotometry, InfraredLarge Language ModelsCitrusMachine Learning

Stichwörter der Autor:innen

Automated chemometricsIn context learningInfrared spectroscopyLarge language modelsSpectral analysis

Autor:innen und Institutionen

  • School of Physics, State Key Laboratory of Optoelectronic Materials and Technologies, Sun Yat-Sen University, Guangzhou, 510275, China.
  • School of Electronics and Communication Engineering, Sun Yat-Sen University, Shenzhen, 518107, China. Electronic address: yuziru@mail.sysu.edu.cn.

Zitieren

APA

Xie, Z., Chen, Z., Liang, J., Yu, X., & Yu, Z. (2026). LUMIR: an LLM-driven unified agent framework for multi-task infrared spectroscopy reasoning. Analytica Chimica Acta, 1382, 344857. https://doi.org/10.1016/j.aca.2025.344857

Vancouver

Xie Z, Chen Z, Liang J, Yu X, Yu Z. LUMIR: an LLM-driven unified agent framework for multi-task infrared spectroscopy reasoning. Analytica Chimica Acta. 2026;1382:344857. doi:10.1016/j.aca.2025.344857

BibTeX
@article{Xie2026,
  author = {Xie, Zujie and Chen, Zixuan and Liang, Jiheng and Yu, Xiangyang and Yu, Ziru},
  title = {LUMIR: an LLM-driven unified agent framework for multi-task infrared spectroscopy reasoning},
  journal = {Analytica Chimica Acta},
  year = {2026},
  volume = {1382},
  pages = {344857},
  doi = {10.1016/j.aca.2025.344857}
}
Kostenlos starten →